Google shipped three Gemini Flash models but held back its flagship Pro

Google shipped three new Gemini models on Tuesday. The flagship 3.5 Pro still isn’t one of them. The company rolled out Gemini 3.6 Flash and Gemini 3.5 Flash-Lite to everyone, gated a cyber-security specialist to governments, and teased a pre-training run for Gemini 4, all while the Pro tier it first promised in May keeps slipping.

What actually shipped

The whole release lives in the Flash tier, which is Google’s cheap, fast workhorse class rather than its frontier model. That framing matters because Google’s flagship Pro line was last updated in February, so a Flash-only launch leaves the top of the range untouched five months on. Developers waiting on a stronger reasoning model got a cheaper mid-tier and a next-generation teaser instead, which is a real answer to “what do I run today” but not the one many were asking. The move also lands mid-cycle against OpenAI and Anthropic, both of which have shipped fresh frontier models this year. Here’s what’s now live.

• Gemini 3.6 Flash is the headline model, and Google pitches it as a “workhorse model” for coding, knowledge work, and multimodal tasks. It uses 17% fewer output tokens than 3.5 Flash and lands at $1.50 per million input tokens and $7.50 per million output, down from $9 on the model it replaces.
• Gemini 3.5 Flash-Lite is the cheap, low-latency option at $0.30 input and $2.50 output per million tokens, pushing 350 output tokens a second for high-throughput jobs like agentic search and document processing.
• Gemini 3.5 Flash Cyber is fine-tuned to find and fix security vulnerabilities. You can’t get it: Google says the model is “exclusively available to governments and trusted partners as part of a limited access pilot program,” per TechCrunch.

3.6 Flash and Flash-Lite are generally available right now through the Gemini API, AI Studio, Android Studio, and Google Antigravity, with Flash-Lite also wired into the Gemini app and Google Search, 9to5Google reports. Both models also went live the same day on Vercel’s AI Gateway, so you can call google/gemini-3.6-flash through the AI SDK without touching Google’s console.

The Pro-shaped hole

Here’s the thread everyone keeps pulling. Back in May, when 3.5 Flash arrived, Google said its Pro version was “already being used internally, and we look forward to rolling it out next month.” That would have meant June. In late June, Business Insider reported the model had slipped to July. July is now nearly over, and Tuesday’s post gave the same non-date: “Gemini 3.5 Pro is currently testing with partners and we plan to make it broadly available as soon as it’s ready.”

Last week Bloomberg reported Google struggled to meet internal performance goals on the model, which reads as a firmer reason than the “more tuning” line the company has stuck to publicly. Meanwhile Google is already looking past it. “We have already started our most ambitious pre-training run yet, for Gemini 4,” the announcement says, a line DevRel lead Logan Kilpatrick amplified alongside the promise that Pro would “land soon.” Shipping a Flash refresh and a next-generation pre-training tease in the same breath, with no dated Pro in between, is an odd shape for a launch.

What this means for you

If you build on Gemini, 3.6 Flash is the practical upgrade to make today. It’s cheaper per token than 3.5 Flash, it burns fewer of them, and it’s the model Google is actually iterating on in the open. For agentic or high-volume work, Flash-Lite at $0.30 input is hard to beat on cost. But don’t architect around 3.5 Pro yet. “Testing with partners” is not a ship date, and a model that has now missed both its June and July windows can miss a third. If your workload needs a genuine frontier tier, Claude Sonnet 5 and the open-weight GLM-5.2 are shipping and benchmarkable while Google’s flagship stays a promise. Watch whether Pro appears before Gemini 4’s pre-training run does.